Copyright 2010 Please consult the authors.eprints.qut.edu.au/32789/1/c32789.pdf · categories for...
Transcript of Copyright 2010 Please consult the authors.eprints.qut.edu.au/32789/1/c32789.pdf · categories for...
QUT Digital Repository: http://eprints.qut.edu.au/
This is the accepted version of this conference paper:
Clothier, Reece A., Palmer, Jennifer L., Walker, Rodney A., & Fulton, Neale L. (2010) Definition of airworthiness categories for civil Unmanned Aircraft Systems (UAS). In: 27th International Congress of the Aeronautical Sciences, 19‐24 September 2010, Acropolis Conference Centre, Nice. (In Press)
© Copyright 2010 Please consult the authors.
27TH
INTERNATIONAL CONGRESS OF THE AERONAUTICAL SCIENCES
1
Abstract
This paper introduces a novel strategy for the
specification of airworthiness certification
categories for civil unmanned aircraft systems
(UAS).
The risk-based approach acknowledges
the fundamental differences between the risk
paradigms of manned and unmanned aviation.
The proposed airworthiness certification matrix
provides a systematic and objective structure
for regulating the airworthiness of a diverse
range of UAS types and operations.
An approach for specifying UAS type
categories is then discussed. An example of the
approach, which includes the novel application
of data-clustering algorithms, is presented to
illustrate the discussion.
1 Introduction
The requirement for regulations governing the
airworthiness of a civil aircraft stems from the
Chicago Convention of 1944 [1]. Article 31 of
the Convention requires aircraft to be
certificated as airworthy, and Article 8 stipulates
the extension of these requirements to UAS. As
described in Annex 8 to the Convention, the
objective of these regulations is to achieve,
“among other things, protection of other
aircraft, third parties and property” [2].
Airworthiness, according to Australian
Defence Force (ADF) instructions [3], is: “…a
concept, the application of which defines the
condition of an aircraft and supplies the basis
for judgement of the suitability for flight of that
aircraft, in that it has been designed,
constructed, maintained and is expected to be
operated to approved standards and limitations,
by competent and approved individuals, who
are acting as members of an approved
organisation and whose work is both certified
as correct and accepted on behalf of the ADF.”
For civil conventionally piloted aviation
(CPA), airworthiness is assured through the
issuance of a Certification of Airworthiness
(CoA) to an individual aircraft or through
special certificates. A CoA is a formal statement
that the aircraft is verified as being compliant
with a prescriptive body of standards and
regulations that are defined based on its type.
The foundation of the CPA airworthiness
regulatory framework is established in Part 21
regulations [4, 5], which prescribe the
applicability of different codes of requirements
to the different types of aircraft and their
intended operation.
International consensus on a prescriptive
framework of airworthiness regulations for civil
unmanned aircraft systems (UAS) has yet to be
reached. Currently, there are no specific
standards and regulations for the type
certification of civil UAS.
In Australia, assurances of the safety of
other airspace users and people and property on
the ground are instead provided by the
placement of restrictions on where UAS
operations may take place. These restrictions are
mandated under Civil Aviation Safety
Regulations (CASR) Part 101 [6], which
specifies that a UAS must be certificated as
airworthy: if the unmanned aircraft1 (UA) has a
maximum takeoff weight (MTOW) above 150
1 The term unmanned aircraft refers only to the airborne
component of a UAS.
DEFINITION OF AIRWORTHINESS CATEGORIES FOR CIVIL UNMANNED AIRCRAFT SYSTEMS (UAS)
Reece A. Clothier,* Jennifer L. Palmer,
† Rodney A. Walker,
* Neale L. Fulton
‡
*Australian Research Centre for Aerospace Automation,
†Defence Science and Technology
Organisation, ‡Commonwealth Scientific and Industrial Research Organisation
Keywords: UAS, UAV, Airworthiness, Civil Regulations, Certification
REECE A. CLOTHIER, JENNIFER L. PALMER, RODNEY A. WALKER, NEALE L. FULTON
2
kg, if the UAS is to be operated over a
‘populated’ area or in controlled airspace, or for
commercial reward. UAS may be certificated in
the experimental designation (e.g., as described
in AC 21-43(0) [7]), for specific applications
and not for commercial reward, but remain
subject to operational restrictions. Operational
regulations such as CASR Part 101 [6] prescribe
the requirement for certification (primarily
based on the nature of the intended operation),
but not the specific type categories or categories
of airworthiness against which a CoA may be
issued.
The absence of a prescriptive airworthiness
certification framework and the subsequent
operational limitations imposed come at
significant expense to the UAS industry. A
prerequisite to the realisation of a viable civil
UAS industry is the definition of an appropriate
airworthiness certification framework for UAS.
This framework must take into consideration the
unique aspects of the technology, their
operations, the market drivers, and the broader
socio-political issues associated with the
integration of a new aviation technology into
society.
2 The Airworthiness Framework
The basis for an airworthiness framework for
civil UAS is established by acknowledging that
the primary entities of value (EoV) are external
to the UAS. The primary hazard of concern is
that of a UAS impacting a region on the ground
and the subsequent losses that could be
registered against different EoV in the region.
UAS may be operated over a diverse range of
areas, and hence the associated degree of risk
varies. Illustrations of the nature of this
dependency are provided in Refs. [8-10].
Conversely, for CPA, the primary risks are to
those people onboard the aircraft and hence
CPA airworthiness regulations are defined “as
far as is practicable” [11] independent of: the
nature of the operational environment and the
purpose for which the aircraft will be used in
service [11]. McGeer and Vagners [12] aptly
encapsulates the differences between the two
airworthiness paradigms:
…with a manned aircraft you have to build
to the same standard no matter what is
underneath you, but among unmanned
aircraft, acceptable safety for flights
exclusively over oceans can be achieved
with rather more rickety machines than
would be fit to fly over a city. [12]
A new concept for the structuring of
airworthiness regulations for civil UAS that
acknowledges this fundamental difference
between paradigms is required. One such
framework is proposed by Clothier et al. [13]
and illustrated in Fig.1. The framework is based
on the principles of a risk matrix and advocated
as being a suitable structure for the definition of
Part 21-equivalent regulations for civil UAS
[13]. The framework is briefly described in the
following sections.
2.1 Operational Environments
The set of m rows of the proposed
airworthiness certification matrix represent the
different environments that a UAS may be
operated over, ranked in order of increasing
‘susceptibility’ of an area to experience loss
given a UAS mishap (i.e., a categorisation of
operational areas ranging from the high seas
through to large open-air gatherings). These
categories are defined independent of the type
of UAS impacting the area. The set of
categories of operational environments must be
disjoint, and provide complete and contiguous
coverage of the range of operational
environments potentially over-flown (i.e.,
provide an unambiguous classification of all
potential areas over which a UAS operation
could occur).
2.2 Type Categories
The set of n columns of the matrix
represent the different type categories of UAS.
Each type category describes a grouping of
UAS where the magnitude of potential loss due
to a mishap is within some pre-defined bounds,
irrespective of where the UAS is operated. Or
more generally, UAS are grouped based on the
question: given the occurrence of an
unrecoverable flight critical failure, what is the
DEFINITION OF AIRWORTHINESS CATEGORIES FOR CIVIL UNMANNED AIRCRAFT SYSTEMS (UAS)
3
Fig. 1 Proposed structure of an airworthiness certification framework for civil UAS [13]
maximum degree of loss the UAS could cause,
irrespective of where it crashed?
The set of type categories must be disjoint
(i.e., provide an unambiguous classification of
the diverse range of UAS) and provide complete
and contiguous coverage of the range of
plausible magnitude of potential loss.
2.3 Operational Scenarios
Each of the m × n cells of the matrix
represents a unique operational scenario,
defined by the combination of a specific UAS
type category with a specific category of
operational environment. For each operational
scenario formed, an assessment of the level of
risk is then made.
2.4 Airworthiness Categories
Illustratively, the assignment of the r
airworthiness categories is the process of
assigning colours to each of the cells of the
matrix illustrated in Fig. 1.
The airworthiness categorisation scheme is
representative of a discrete, contiguous and
increasing ranking of risk (e.g., MIL-STD-882D
defines the risk-ranking scheme of low,
medium, serious, and high [14]). Each
operational scenario may then be mapped to one
of the airworthiness certification categories
based on the levels of risk assessed for the
operational scenario. Operational scenarios with
‘similar’ levels of risk may logically be assigned
to the same airworthiness certification category.
In general, the operational scenarios (and
subsequently the certification categories
assigned to them) in the lower-right quadrant of
the matrix present higher levels of risk than are
associated with the scenarios in the upper-left
quadrant.
Unlike the CPA airworthiness framework,
the aircraft-type category does not directly
define the airworthiness category. A UAS of a
specific type may be certificated in one or more
airworthiness categories dependent on the
category of operational area over which it is
intended to be flown. The approach permits
UAS manufacturers to develop and certify a
UAS-type for a specific application, operational
environment or price-point in the market.
REECE A. CLOTHIER, JENNIFER L. PALMER, RODNEY A. WALKER, NEALE L. FULTON
4
A significant challenge is the determination
of an appropriate number of airworthiness
certification categories. The minimum number
of categories is one. This represents the
undesirable case where all UAS operational
scenarios are regulated under the same
standards irrespective of the differences in the
associated levels of risk.
The maximum number of airworthiness
categories is equal to the number of rows, m,
multiplied by the number of columns, n. This
represents the case where a separate body of
airworthiness regulations is defined for each of
the operational scenarios. This case offers the
maximum possible degree of tailoring of
regulations. However, Clothier et al. [13]
identify several disadvantages to a large number
of airworthiness categories. A determination of
the optimum number of airworthiness categories
requires subjective trade-offs to be made, for
example, determining the number of
airworthiness categories necessary to:
1. minimise the relative differences
between the levels of risk associated
with each operational scenario assigned
to a particular certification category (i.e.,
ensure operational scenarios assigned to
a given certification category are indeed
comparable in their risks);
2. ensure sufficient resolution in the
certification categories to permit niche or
unique operational scenarios; and
3. ensure a practical and workable
categorisation from the perspective of
the authority charged with promulgation
of the regulation (e.g., not unmanageable
in number).
Given an appropriate assignment of
certification categories to operational scenarios,
regulations may then be developed (or existing
CPA regulations tailored and adopted) to each
scenario. An example of this process, for civil
UAS Part 1309 regulations, is described in [13].
2.5 Summary
The airworthiness certification matrix
provides a flexible, yet systematic and
defensible framework for regulating the
airworthiness of civil UAS. The matrix provides
a suitable structure for simplifying the
classification of the diversity of systems and
operations. A systematic and transparent
process for assigning airworthiness categories is
described based on assessments of risk. This
assignment considers both the system and the
environment, hence the proposed airworthiness
certification matrix acknowledges the
fundamental differences between the CPA and
the UAS risk paradigms.
It is for these reasons, and others, that
Clothier et al. [13] advocate the proposed
airworthiness certification matrix as a suitable
basis for defining a Part 21-equivalent
regulation for civil UAS.
The remainder of this paper describes a
possible strategy for the specification of a
principal component of the proposed
airworthiness certification matrix, that of UAS
type categories (i.e., the columns of Fig. 1).
3 Specification of UAS Type Categories
For a comprehensive review of existing UAS
type categorisation schemes refer to the
forthcoming work of Nas [15].
Currently, there is no consensus on the
definition of type categories for UAS [16, 17],
with many of the existing schemes having not
been defined for the purpose of certification. In
addition, there is no consensus on the process
for defining a suitable scheme when such
consensus is fundamental to the progress of
regulations [16]. The specification of type
categories could have a significant influence on
the future ‘shape’ of the civil UAS industry,
hence a more objective and systematic process
for defining UAS type categories is needed. The
following sub-section describes one possible
approach that could be used to define type
categories for UAS.
3.1 A Risk-Based Approach
A type category is a grouping of UAS that
are similar in some relevant way. In accordance
with the structure of the proposed airworthiness
matrix (Fig. 1), the measure of similarity used to
DEFINITION OF AIRWORTHINESS CATEGORIES FOR CIVIL UNMANNED AIRCRAFT SYSTEMS (UAS)
5
define UAS type categories is the theoretical
magnitude of potential loss due to a mishap
(§2.2). A measurement of loss may be made for
a range of different types of EoV (e.g., people,
property, and environment) and entity attributes
(e.g., for people: physical, psychological, and
financial). Within the context of defining
airworthiness regulations for civil UAS, the
primary loss of concern is the magnitude of
physical harm to third parties on the ground.
Proceeding on this premise, a generalised two-
dimensional (2D) space describing the
magnitude of potential loss to people on the
ground is defined in Fig. 2. The two component-
dimensions are defined as:
X (or horizontal axis): the types of
potential harm to people exposed to a
mishap; and
Y (or vertical axis): the maximum number
of people that could be harmed given a
mishap.
Fig. 2 Illustration of the assignment of UAS type
categories within a 2D space describing the magnitude of
potential loss to people on the ground due to a UAS
mishap
The first component-dimension aims to
distinguish between UAS types based on their
ability to cause levels of physical harm to
people exposed to a UAS mishap. For example,
the physical and aerodynamic properties of
some UAS make them very unlikely to cause a
fatal injury to a person struck in the open. On
the other hand, some UAS are capable of
penetrating the strongest of structures and
consequently have the potential to cause harm to
the people sheltered within buildings.
A range of different hazards must be
considered to adequately characterise the ability
of a UAS to cause harm to people on the
ground, including:
1. primary hazards – e.g., the transfer of
energy through a direct strike, flying
debris, heat from an explosion, and
incident pressure waves; and
2. secondary hazards – e.g., the ensuing
collapse of a building, bush fires, and
on-going contamination by hazardous
substances.
To simplify this example, for this paper, the
UA is assumed to be in-frangible, and the UA
and any object that it strikes (e.g., a building or
vehicle) are inert (e.g., no secondary explosions
or collapsing walls are assumed possible);
however, a more comprehensive approach
would include a broader range of hazards.
Based on these assumptions, the primary hazard
may be considered as the transfer of kinetic
energy from the UA to people and structures on
the ground directly struck by the UA. The
measure used to characterise this component-
dimension is the maximum possible kinetic
energy of the UA on impact with the ground
(KEmax), given by:
2
maxMV2
1=maxKE (1)
where M is the MTOW of the UA and Vmax is its
maximum speed.
The second component-dimension aims to
distinguish amongst UAS types based on the
maximum number of potential casualties due to
mishap. Such a distinction is necessary to reflect
society’s heightened apprehension towards
mishaps with a large magnitude of potential
loss, irrespective of the associated likelihood of
occurrence [18].
Assuming independence of the particular
region over-flown (i.e., assuming an exposed
population of uniform distribution and of
REECE A. CLOTHIER, JENNIFER L. PALMER, RODNEY A. WALKER, NEALE L. FULTON
6
uniform susceptibility or response to an incident
stress), the potential number of casualties is a
complex function of the spatial-temporal
distribution of the effects of a mishap. This
spatial-temporal region is referred to as the
hazard area.
Within the context of the simplified
example, the hazard area is the impact area of
the crashing UA. A range of models describing
the impact area of a crashing aircraft are
described in Refs. [19-24]. In this example, the
impact area, Iarea, is modelled as a simplified
glide area for a fixed-wing UA [21, 24], given
by:
,)2++(×)2+(= pglidepspanarea RDLRWI
(2)
where Wspan is the wingspan, Rp is the average
radius of a person (the shoulder width of an
average person), L is the length of the aircraft,
and Dglide is the distance along the ground
travelled by the UA at glide path angle, γ, from
the height of an average person, Hp, given by:
.tan γ
pglide
HD = (3)
3.1.1 Defining the Type Categories
The process of defining type categories is
one of determining mutually exclusive regions
within the 2D space where the magnitude of
potential loss due to a UAS mishap is similar
(Fig. 2). Two approaches for guiding the
specification of these regions are investigated in
this paper:
1. definition of limits with respect to one or
both of the component-dimensions (i.e.,
specification of loss criteria); and
2. application of a data-clustering
algorithm to ‘learn’ type categories from
a UAS dataset.
To ensure an unambiguous classification of
UAS types, the specification of type categories
must also satisfy:
PLoss(i) < PLoss(i + 1), for 1 ≤ i < r, (4)
where PLoss(i) is the potential magnitude of loss
associated with the ith
type category. Satisfying
this relationship requires subjective judgements
on the relative significance of the two
components used to represent the magnitude of
potential loss (i.e., the ranking of different loss
outcomes).
3.1.2 Categorisation Using Limits
The first approach to defining UAS type
categories is to specify boundaries with respect
to one or both axes of the 2D space describing
the magnitude of potential loss due to a UAS
mishap.
To illustrate, four high-level type
categories are defined based on the energy
required to cause a particular level of harm to a
person exposed to a UAS mishap (i.e., along the
impact-energy component-dimension). The type
categories and associated impact-energy
boundaries are summarised in Table 1. The first
category describes UAS that have sufficient
impact energy to cause injury (but not a fatal
injury) to one or more people exposed in the
open. A range of different injury scales could be
used to further describe this category (e.g., the
Abbreviated Injury Scale in Ref. [25]). The
second category describes UAS with impact
energy greater than 42 J but less than 1,356 J,
which are capable of causing fatal injuries to
one or more unsheltered people. The third
category describes UAS with sufficient impact
energy to penetrate a typical residential structure
(i.e., a corrugated-iron roof). The final category
describes UAS that are capable of penetrating
the highest level of protection afforded to the
general public (i.e., a reinforced concrete
structure).
A database2 comprised of over 500 rotary
and fixed-wing UAS types is used to illustrate
the outcome type categories. Three-hundred and
eighty-three fixed-wing UAS types are mapped
to the 2D space by use of Eqs. 1–3. These UA
are then colour coded based on the type-
classification scheme described above. Fig. 3
shows the classification of the fixed-wing UAS
dataset based on the categories defined in Table
1. Fig. 4 shows the same classification mapped
2 Database of UAS compiled and maintained by Defence
Science and Technology Organisation (DSTO) personnel.
Database includes military UAS.
DEFINITION OF AIRWORTHINESS CATEGORIES FOR CIVIL UNMANNED AIRCRAFT SYSTEMS (UAS)
7
Category Description of loss outcome Energy Limit* (J) Description of Energy Limit
1 UAS capable of causing a non-fatal
injury to one or more exposed
people
KEmax < 42
< 5% probability of causing a fatal injury
to an individual standing in the open [26]
2 UAS capable of causing a fatal
injury to one or more exposed
people
42 ≤ KEmax < 1,356
≥ 5% probability of causing a fatal injury
to an individual standing in the open [26]
3 UAS capable of causing a fatal†
injury to one or more people within
a typical residential structure
1,356 ≤ KEmax < 13,560
Capable of penetrating a corrugated-iron
roof house [27]
4 UAS capable of causing a fatal†
injury to one or more people within
a typical commercial structure
KEmax ≥ 13,560
Capable of penetrating a reinforced
concrete structure [27]
*Energy limits are based on the conservative assumption that each UA may be represented as an in-frangible piece of
inert debris. †It is assumed that any individual inside a structure is fatally injured if the UA has sufficient energy to penetrate the
structure.
Table 1 Type categories based on the ability of the UAS to cause harm
Fig. 3 Type categorisation of fixed-wing UAS based on
impact energy limits
Fig. 4 Type categorisation displayed using UA MTOW
against the MTOW of the UA, a common metric
used in existing type categorisation schemes.
The mapping of the categories with respect
to the MTOW of each UA type, shown in Fig. 4,
results in an ambiguous classification. This
highlights that a measure of the MTOW of a
UA, on its own, may not provide a good
discriminator with respect to the potential harm
it may cause. One must also consider the
maximum potential speed of the UA upon
impact. From Fig. 3 and 4, it may be observed
most UAS of MTOW greater than 200 g are
capable of inflicting a fatal injury to one or
more people exposed in the open. In addition,
nearly all UAS with a MTOW greater than ~20
kg have sufficient maximum impact energy to
penetrate the highest level of protection
typically available to a member of the general
public. Consequently, type category 4 covers a
disproportionately large range of UAS types.
Greater resolution within the type categories
may be required. This may be obtained by
defining additional categories of harm along the
impact-energy axis (e.g., types of sheltering,
levels of injury, or a combination of both) or by
defining limits with respect to the second
component-dimension (i.e., the impact area).
Classifying with respect to the maximum
impact area of the UA attempts to distinguish
UAS type categories based on the number of
REECE A. CLOTHIER, JENNIFER L. PALMER, RODNEY A. WALKER, NEALE L. FULTON
8
people potentially harmed. The larger the
impact area, the more people potentially
exposed to harm given a UAS mishap. Relating
the impact area to a number of potential
casualties (i.e., through the specification of a
population distribution) and the specification of
category boundaries in terms of a number of
casualties is a subjective process. A more
objective approach is to use a data-clustering
algorithm to objectively ‘learn’ type categories
from the UAS dataset.
3.1.3 Categorisation by Use of Data Clustering
Data clustering or data segmentation [28]
is a process of unsupervised learning [29] that
attempts to organise a collection of objects into
a finite number of meaningful groupings, or
clusters, based on some measure of similarity
[30]. In this example, the objects are individual
UAS types and the measure of similarity is the
impact area of the UA, given by Eq. 2.
Data-clustering techniques have been
widely applied, particularly in the field of image
processing, with clustering algorithms being
used for image compression, segmentation, and
pattern recognition for machine vision. It is not
the objective of this paper to provide a thorough
overview of data-clustering techniques. For this,
the reader is referred to the introductory paper
of Jain and Dubes [31]. Instead, the aim here is
to emphasise that an alternative and more
objective approach to the definition of UAS
type-certification categories may lie in the
application of data clustering techniques.
In this example, the widely applied [32] k-
means algorithm [33] is used to determine sub-
type categories within the fourth type category
illustrated in Fig. 3. The k-means algorithm is a
partitioning relocation algorithm [32] that
determines a finite number of k clusters/groups
directly from the data by iteratively relocating
points until a minimum in an objective distance
measure is obtained. k is an integer value less
than or equal to the number of unique UAS
types in the dataset (K).
The k-means algorithm has the advantages
of being extensively used, simple to interpret
and implement, and computationally efficient.
One of the significant disadvantages is that the
convergence of the algorithm is highly
dependent on its initialisation state [28-30].
Dependent on the initial conditions, the
algorithm may converge to a local optimum [28]
and consequently may not find a globally
optimal partitioning of UAS types into
categories. Another significant disadvantage of
the algorithm is that the number of clusters (k)
must be known a priori.
To address these disadvantages, the k-
means algorithm was run for values of k ranging
from 2 to 7. The algorithm was run fifty times
for each value of k, with randomly selected
initial conditions. The solution with the lowest
sum of intra-cluster distances was then selected
from the set of fifty possible solutions. As
advocated by Kaufman and Rousseeuw [32],
the silhouette coefficient (SC) may be used to
objectively compare the performance of the
clustering algorithm for each value of k. SC is
the maximum of the average silhouette width
for the entire dataset, a dimensionless measure
of how well each object within a dataset has
been classified [32] (refer to p.83 of Ref. [32]
for a description of its calculation). SC ranges
between –1 and +1, with the values of –1, 0 and
+1 indicative of a poor, an indifferent, and an
appropriate classification of UAS into type
categories, respectively. SC for each value of k
is presented in Table 2.
No. of type
sub-
categories,
k
Maximum
average SC
Qualitative assessment* of
how well the classification
represents the structure of
the dataset
2 0.822 strong
3 0.705 strong
4 0.706 strong
5 0.692 reasonable
6 0.710 strong
7 0.720 strong *Based on the qualitative assessment table presented in
Kaufman and Rousseeuw [32].
Table 2 Results obtained from clustering approach for
different numbers of type categories
Based on the results presented in Table 2,
the mathematically optimal solution is to sub-
divide the fourth UAS type category into two
sub-categories (k = 2, Table 2). This results in a
total of five UAS type categories.
DEFINITION OF AIRWORTHINESS CATEGORIES FOR UNMANNED AIRCRAFT SYSTEMS (UAS)
9
Fig. 5 Classification of UAS into five type categories
Type
category
Boundary conditions Example UAS
1 KEmax < 42 J Black Widow,
Hornet
2 42 J ≤ KEmax < 1,356 J Pointer, Raven
3 1,356 J ≤ KEmax < 13,560 J ScanEagle,
Aerosonde Mk4
4 13,560 J ≤ KEmax
Iarea < 347 m2
Shadow 600
5 347 m2 ≤ Iarea
Heron 1,
Taranis, Global
Hawk
Table 3 Example output UAS type categorisation scheme
A visualisation of the candidate five-category
UAS type categorisation scheme is illustrated in
Fig. 5; and Table 3 summarises the boundaries
of the example type categorisation scheme.
It may be observed in Table 2 that the SC
determined for each value of k varies little, and
hence the ideal number of type categories is
more likely to be determined by the context of
the problem (i.e., broader subjective trade-offs,
as described in §2.4), rather than by the
mathematically optimal solution. Thus, the
results presented in Fig. 5 and Table 3 are
illustrative only.
3.4 Discussion
The assumptions and simplified model
used in the previous example place limitations
on the usability of the results; however, the
intention here is to illustrate an objective
approach for specifying UAS type categories.
The basis for classification is the magnitude of
potential loss a UAS may cause given a mishap,
REECE A. CLOTHIER, JENNIFER L. PALMER, RODNEY A. WALKER, NEALE L. FULTON
10
independent of where it is flown. Two
techniques for specifying the boundaries of the
type categories have been described. However,
it is unlikely that this process will ever be
reduced to a purely mathematical approach. The
specification of type categories will heavily
influence the nature of the future
civil/commercial UAS industry; hence, it is
likely the decision-making process will involve
multiple, competing stakeholders and be a
predominantly discursive process. The
approaches presented in this paper provide a
focus for a risk-informed basis to help guide this
higher-order decision-making process.
4 Summary and Conclusions
This paper summarises a novel strategy for the
specification of airworthiness certification
categories for civil UAS. The proposed
framework, outlined in §2, comprises two
orthogonal dimensions (representative of the
potential magnitude of loss and the potential for
realising loss, given a UAS impacting the
ground). The Cartesian product of the two
dimensions defines a finite set of operational
scenarios, to which airworthiness categories are
then assigned based on an assessment of the
risks. In §3, the specification of the columns of
the matrix (a type-categorisation scheme for
UAS) was discussed and presented in simplified
form. As discussed in §2.4, an appropriate
system of certification categories for the
purpose of regulating civil UAS may be based
on such a risk matrix.
The risk matrix approach permits a
tailoring of airworthiness regulation in line with
the risks. For example, small UAS, which
present lower levels of risk, are subject to lower
standards of airworthiness and hence may be
permitted a higher mishap rate. Larger UAS,
which present risks comparable to CPA, attract
a level of airworthiness regulation comparable
to CPA.
The airworthiness matrix approach
provides a systematic and justifiable (through
traceability in risk) framework for the regulation
of the diverse range of UAS and their
operations. The matrix affords a greater degree
of flexibility in the regulation of UAS allowing
UAS manufacturers to develop and certify a
UAS-type for a specific application, operational
environment or price-point in the market.
Broader social, political, technical, and
practical considerations will ultimately
determine the refinement of any airworthiness
certification scheme adopted for civil UAS;
however, the approach presented in this paper
provides the risk-informed foundations from
which to guide these discussions. The
comprehensive specification of the
airworthiness certification matrix, including the
specification of UAS type categories and
operational environments, will be the subject of
a future paper.
5 Contact Author Email Address
The corresponding author for this paper is Mr.
Reece Clothier: [email protected].
Acknowledgement
The authors would like to thank Mr. Michael
Nas, from Murdoch University, Perth, Australia,
and members from the Australian Aerospace
Industry Forum, Certification and Regulation
Working Group, Unmanned Aircraft System
Sub-Committee for their input to this paper. The
authors would also like to thank Dr XunGuo Lin
for his comments made in reviewing this paper.
This research is supported by a Queensland
State Government Smart State PhD Scholarship.
References
[1] ICAO. Convention on international civil
aviation, eighth edition. International Civil
Aviation Organization (ICAO), 2000.
[2] ICAO. Annex 8 to the convention on
international civil aviation, airworthiness of
aircraft (10th edition). International Civil
Aviation Organization (ICAO), April, 2005.
[3] ADF. Defence instruction (general) ops 02–2,
Australian Defence Force airworthiness
management. Department of Defence, Canberra,
ACT, 2002.
[4] CASA. Civil aviation safety regulations 1998
(CASR) part 21, certification and airworthiness
requirements for aircraft and parts. Civil
Aviation Safety Authority (CASA), Australia,
ACT, Canberra, 2003.
DEFINITION OF AIRWORTHINESS CATEGORIES FOR CIVIL UNMANNED AIRCRAFT SYSTEMS (UAS)
11
[5] FAR. Code of federal regulations, Title 14,
federal aviation regulation, part 21, certification
procedures for products and parts Washington,
DC, USA, 2009.
[6] CASA. Civil aviation safety regulations 1998
(CASR) part 101, unmanned aircraft and rocket
operations. Civil Aviation Safety Authority,
Canberra, ACT, Australia, December, 2004.
[7] CASA. AC 21-43(0), Experimental certificate
for large unmanned aerial vehicle (UAV). Civil
Aviation Safety Authority (CASA), Canberra,
ACT, Australia, June, 2006.
[8] R. Clothier and R. Walker. Determination and
evaluation of UAV safety objectives. In 21st
International Unmanned Air Vehicle Systems
(UAVS) Conference, Bristol, UK, 2006.
[9] R. Weibel and R. Hansman. Safety
considerations for operation of different classes
of UAVS in the NAS (AIAA-2004-6421). In
AIAA 3rd Unmanned Unlimited Technical
Conference, Workshop and Exhibit, Chicago, IL,
USA, 2004.
[10] K. Dalamagkidis, K. P. Valavanis and L. A.
Piegl. On integrating unmanned aircraft systems
into the national airspace system. Springer,
2009.
[11] D. R. Haddon and C. J. Whittaker. Aircraft
airworthiness standards for civil UAVs. Civil
Aviation Authority (CAA), London, UK, 2002.
[12] T. McGeer and J. Vagners. Wide-scale use of
long-range miniature Aerosondes over the
world's oceans. Insitu Group, Bingen, WA,
USA, 1999.
[13] R. A. Clothier, J. L. Palmer, R. A. Walker, and
N. L. Fulton. Definition of an airworthiness
certification framework for civil unmanned
aircraft systems. Submitted to Safety Science,
2010.
[14] DoD. MIL-STD-882D, Standard practice for
system safety. United States of America
Department of Defense, Washington, DC, USA,
10 February, 2000.
[15] M. Nas. Classifying unmanned aircraft systems:
Development of an objective framework for
evaluating uas classification schemes. Murdoch
University, Perth, WA, Australia, 2010.
[16] M. T. DeGarmo. Issues concerning integration
of unmanned aerial vehicles in civil airspace.
MITRE, Center for Advanced Aviation System
Development, McLean, VA, USA, November,
2004.
[17] K. Dalamagkidis, K. P. Valavanis and L. A.
Piegl. A survey of unmanned aircraft systems
regulation: Status and future perspectives. In
16th Mediterranean Conference on Control and
Automation, Ajaccio, France, pp. 717–723, 2008.
[18] M. E. T. Horn, N. L. Fulton and M. Westcott.
Measures of societal risk, and their potential use
in civil aviation. Risk Analysis, Vol. 28, No. 6,
pp. 1711–1726, 2008.
[19] M. Lin, E. Larson and J. Collins. Columbia
accident investigation board report Volume ii,
Appendix D.16, Determination of debris risk to
the public due to the Columbia breakup during
reentry. Columbia Accident Investigation Board,
Arlington, VA, USA, 2003.
[20] R. M. Montgomery and J. A. Ward. Casualty
areas from impacting inert debris for people in
the open. Research Triangle Institute, Cocoa
Beach, FL, USA, 1995.
[21] F. Grimsley. Equivalent safety analysis using
casualty expectation approach (AIAA-2004-
6428). In AIAA 3rd Unmanned Unlimited
Technical Conference, Workshop and Exhibit,
Chicago, IL, USA, 2004.
[22] FAA. Advisory Circular AC431.35-1, Expected
casualty calculations for commercial space
launch and reentry missions. U.S. Department of
Transportation, Federal Aviation Administration,
Washington, DC, USA, 2000.
[23] R. Clothier, R. Walker, N. Fulton, and D.
Campbell. A casualty risk analysis for unmanned
aerial system (UAS) operations over inhabited
areas. In Twelfth Australian International
Aerospace Congress, 2nd Australasian
Unmanned Vehicles Conference, Melbourne,
VIC, Australia, 2007.
[24] RCC. Range safety criteria for unmanned air
vehicles, rationale and methodology supplement
to 323-99. Range Commanders Council, Range
Safety Group, NM, USA, 2001.
[25] AAAM. Abbreviated injury scale (AIS) 2005
manual, update 2008. 1st edition. Association
for the Advancement of Automotive Medicine,
Chicago, IL, USA, 2005.
[26] RCC. Standard 321-02, Common risk criteria for
national test ranges: Inert debris. Range Safety
Group Risk Committee, Range Commanders
Council, US Army White Sands Missile Range,
NM, USA, June, 2002.
[27] Risk-based explosives safety analysis, Technical
Paper No. 14. US Department of Defense,
Explosives Safety Board, Alexandria, VA, USA,
February, 2000.
[28] J. Han and M. Kamber. Data mining: Concepts
and techniques. 2nd edition. Morgan Kaufman,
San Francisco, CA, USA, 2006.
[29] P. Berkhin. A survey of clustering data mining
techniques. In Grouping multidimensional data:
Recent advances in clustering, Springer Berlin
Heidelberg, New York, NY, USA, 2006.
[30] A. K. Jain, M. N. Murty and P. J. Flynn. Data
clustering: A review. ACM Computing Surveys,
Vol. 31, pp. 264–323, 1999.
[31] A. K. Jain and R. C. Dubes. Algorithms for
clustering data. Prentice-Hall Inc, Upper Saddle
River, NJ, USA, 1988.
[32] L. Kaufman and P. J. Rousseeuw. Finding
groups in data: An introduction to cluster
analysis. Wiley, New York, NY, USA, 1990.
REECE A. CLOTHIER, JENNIFER L. PALMER, RODNEY A. WALKER, NEALE L. FULTON
12
[33] J. MacQueen. Some methods for classification
and analysis of multivariate observations. Fifth
Berkeley Symposium on Mathematical Statistics
and Probability, pp. 281–297, 1967.
Copyright Statement
The authors confirm that they, and/or their
company or organization, hold copyright on all
of the original material included in this paper.
The authors also confirm that they have
obtained permission, from the copyright holder
of any third party material included in this
paper, to publish it as part of their paper. The
authors confirm that they give permission, or
have obtained permission from the copyright
holder of this paper, for the publication and
distribution of this paper as part of the
ICAS2010 proceedings or as individual off-
prints from the proceedings.